#jogamp @ irc.freenode.net - 20131217 05:05:33 (UTC)


20131217 05:05:33 -jogamp- Previous @ http://jogamp.org/log/irc/jogamp_20131216050533.html
20131217 05:05:33 -jogamp- This channel is logged @ http://jogamp.org/log/irc/jogamp_20131217050533.html
20131217 08:05:13 * eclesia (~husky@anon) has joined #jogamp
20131217 08:18:27 * monsieur_max (~maxime@anon) has joined #jogamp
20131217 08:26:06 * xranby_ac100 (~xranby@anon) Quit (Ping timeout: 246 seconds)
20131217 09:08:07 * xranby (~xranby@anon) Quit (Ping timeout: 260 seconds)
20131217 09:10:11 * xranby (~xranby@anon) has joined #jogamp
20131217 09:23:20 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20131217 09:29:29 * monsieur_max (~maxime@anon) has joined #jogamp
20131217 09:48:21 * xranby (~xranby@anon) Quit (Ping timeout: 246 seconds)
20131217 09:52:58 * xranby (~xranby@anon) has joined #jogamp
20131217 11:58:31 <xranby> https://github.com/offbytwo/git-hg <- handy if we want to clone icedtea-web hg into a git repository
20131217 11:59:37 <xranby> it allows expotring any git changes back to hg as well
20131217 12:04:18 <xranby> yesterday i ran some tests using the GLMediaPlayer on raspberry pi to test audio only playback.
20131217 12:05:39 <xranby> initialization is still quite heavy.. for some reason gl initialization happen during GLMediaPlayerFactory.getDefault()
20131217 12:06:01 <xranby> that single code line takes about 15s to complete on the pi
20131217 12:06:49 <xranby> after initialization is done playback work ok with about 12% CPU use to playback an mp3 file
20131217 12:07:33 <xranby> playback is smooth and no stuttering is observed while using the pi system while the mp3 playback is running
20131217 12:10:08 <xranby> its possible that we can postpone psm = new GLPixelStorageModes(); inside FFMPEGMediaPlayer to be created only for vid != -2 streams
20131217 12:10:30 <xranby> to remove the GL init for audio only streams
20131217 12:11:06 <sgothel> hi Xerxes
20131217 12:11:16 <xranby> sgothel: good morning
20131217 12:11:22 <sgothel> GL init on pi ~ once 15s .. I see
20131217 12:11:43 <sgothel> ofc
20131217 12:11:52 <xranby> not sure what takes the most time, unpacking zip files or enumerating gl
20131217 12:12:08 <sgothel> AFAIK .. it's the actual GL instance creation w/ EGL ..
20131217 12:12:16 <sgothel> but only the very first time ..
20131217 12:12:25 <sgothel> lib loading, DSP programming .. and all that ..
20131217 12:12:38 <xranby> yes it is a hit at application startup
20131217 12:12:49 <sgothel> 2nd GLContext is not so expensive ..
20131217 12:13:12 <sgothel> so only of you don't trigger GLProfile/JOGL init .. you could avoid it ..
20131217 12:13:19 <sgothel> I dunno :)
20131217 12:13:23 <xranby> we may work on skipping the initialization for audio only streams
20131217 12:13:30 <xranby> yes i will look into it
20131217 12:13:43 <sgothel> you know .. the !GL path in a GL-tool is 'weird' :)
20131217 12:15:30 <xranby> it is possibly a racecondition in when doing audio only playback
20131217 12:15:46 <xranby> when i use the raspimusic code submitted by the user
20131217 12:16:05 <xranby> the application quits sometimes after initializing the stream
20131217 12:16:11 <xranby> and sometimes it playback
20131217 12:16:23 <xranby> thus the application get the wrong mediaplayer event
20131217 12:16:26 <xranby> or something
20131217 12:16:31 <xranby> i need to investigating it further
20131217 12:18:40 <xranby> triggered using this MediaPlayer (audio only) src https://github.com/dudewithasock/raspimusic/blob/master/raspberry/src/MediaPlayer.java
20131217 12:23:56 <sgothel> the GLMediaPlayer listener shall wait until Initialized, the start playing - and wait until EOS
20131217 12:24:15 <sgothel> if EOS happens too early .. maybe a bug
20131217 12:24:38 <xranby> maybe a bug in ffmpeg
20131217 12:24:40 <xranby> ?
20131217 12:24:43 <sgothel> i.e. I introduced some 'manual' EOS detection in last patch
20131217 12:24:59 <sgothel> you see it w/ the DEBUG flags on ..
20131217 12:25:49 <sgothel> -Djogl.debug.GLMediaPlayer
20131217 12:25:54 <sgothel> -Djogl.debug.GLMediaPlayer -Djogl.debug.GLMediaPlayer.Native
20131217 12:28:55 <xranby> kudos
20131217 12:29:39 <xranby> on my desktop system GLMediaPlayerFactory.createDefault() takes only 579ms
20131217 12:33:00 <sgothel> it loads the native libs and performs: FFMPEGMediaPlayer.isAvailable() ..
20131217 13:28:18 <xranby> on my ac100 GLMediaPlayerFactory.createDefault() takes 3112ms
20131217 13:31:25 <xranby> 2229ms using jamvm
20131217 13:54:49 <xranby> FFMPEGDynamicLibraryBundleInfo static block contains GLProfile.initSingleton();
20131217 13:54:53 <xranby> i wonder why
20131217 14:20:54 <xranby> // native ffmpeg media player implementation is included in jogl_desktop and jogl_mobile
20131217 14:20:56 <xranby> thats why
20131217 14:50:31 <xranby> not much to do except checking where time is spent during GLProfile.initSingletion()
20131217 15:00:12 <xranby> on x86: 805ms time spent do do GLMediaPlayerFactory.createDefault(); is roughly divided into:
20131217 15:00:12 <xranby> 132ms NativeWindowFactory.initSingleton();
20131217 15:00:12 <xranby> 100ms GLDrawableFactory.initSingleton();
20131217 15:00:12 <xranby> 71ms initProfilesForDevice(defaultEGLDevice)
20131217 15:00:12 <xranby> 312ms initProfilesForDevice(defaultDesktopDevice)
20131217 15:32:22 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20131217 15:54:48 * xranby (~xranby@anon) Quit (Ping timeout: 246 seconds)
20131217 16:07:48 * monsieur_max (~maxime@anon) has joined #jogamp
20131217 16:48:37 * eclesia (~husky@anon) Quit (Ping timeout: 248 seconds)
20131217 16:56:55 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20131217 18:27:47 * monsieur_max (~maxime@anon) has joined #jogamp
20131217 18:34:17 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20131217 18:38:10 * monsieur_max (~maxime@anon) has joined #jogamp
20131217 19:23:21 <sgothel> @Xerxes: Desktop: GLProfile.initSingleton() ~ 200-300ms
20131217 19:24:10 <sgothel> Use 'TestShutdownCompleteNEWT' to measure 'bringup' performance
20131217 20:54:02 * xranby_ac100 (~xranby@anon) has joined #jogamp
20131217 21:06:56 <xranby_ac100> on ARM the second run of GLProfile.initSingleton is ok and only take ~ 290-350ms tested with TestShutdownCompleteNEWT
20131217 21:08:21 <xranby_ac100> its the first run test01OneLife that needs some crunching ~2500ms on my ARM ac100
20131217 21:09:18 <xranby_ac100> on the pi i was able to lower the startup cost from 12 to 7 seconds by picing atomic jars. thus skipping enumerating the desktop gl
20131217 21:34:18 * 65MAAFL6X (~gorgous@anon) has joined #jogamp
20131217 21:39:39 <xranby_ac100> good evening
20131217 21:43:33 * monsieur_max (~maxime@anon) has left #jogamp
20131217 21:46:03 * 65MAAFL6X (~gorgous@anon) Quit (Ping timeout: 250 seconds)
20131217 22:01:18 <sgothel> @Xerxes: Maybe the mobile_all (sic!) will do the same, i.e. no desktop GL ..
20131217 22:01:26 <sgothel> sounds good ..
20131217 22:05:56 * monsieur_max (~maxime@anon) has joined #jogamp
20131217 22:28:38 <xranby_ac100> sgothel: yup jogl-all-mobile.jar initializes quicker as well
20131217 22:44:51 * monsieur_max (~maxime@anon) has left #jogamp
20131217 22:52:25 * xranby_ac100 (~xranby@anon) Quit (Remote host closed the connection)
20131217 23:23:11 * [Mike] (~Mike]@anon) Quit (Ping timeout: 272 seconds)
20131218 05:01:15 * kermyt (~kermyt@anon) Quit (Ping timeout: 252 seconds)
20131218 05:03:42 * kermyt (~kermyt@anon) has joined #jogamp
20131218 05:05:33 -jogamp- Continue @ http://jogamp.org/log/irc/jogamp_20131218050533.html